paircomp: Data Structure for Paired Comparisons

Description

A class for representing data from paired comparison experiments along with methods for many generic functions.

Usage

paircomp(data,
    labels = NULL, mscale = NULL, ordered = FALSE, covariates = NULL)

Value

paircomp returns an object of class "paircomp" which is a matrix (essentially data) with all remaining arguments of paircomp as attributes (after being checked and potentially suitably coerced or transformed).

Arguments

data: matrix. A matrix with integer values where the rows correspond to subjects and the columns to paired comparisons between objects. See below for details.
labels: character. A vector of character labels for the objects. By default a suitable number of letters is used.
mscale: integer. A vector of integers giving the measurement scale. See below for details. By default guessed from data.
ordered: logical. Does data contain both orderings of each comparison?
covariates: data.frame. An optional data.frame with object covariates, i.e., it must have the same number of rows as the length of labels. May be NULL (default).

Details

paircomp is designed for holding paired comparisons of \(k\) objects measured for \(n\) subjects.

The comparisons should be coded in an integer matrix data with \(n\) rows (subjects) and \(k \choose 2\) columns (unless ordered = TRUE, see below). The columns must be ordered so that objects are sequentially compared with all previous objects, i.e.: 1:2, 1:3, 2:3, 1:4, 2:4, 3:4, etc. Each column represents the results of a comparison for two particular objects. Positive values signal that the first object was preferred, negative values that the second was preferred, zero signals no preference. Larger absolute values signal stronger preference.

mscale provides the underlying measurement scale. It must be a symmetric sequence of integers of type (-i):i where i must be at least 1. However, it may exclude 0 (i.e., forced choice).

If ordered = TRUE, the order of comparison matters and thus data is assumed to have twice as many columns. The second half of columns then corresponds to the comparisons 2:1, 3:1, 3:2, 4:1, 4:2, 4:3, etc.

Examples

Run this code

## a simple paired comparison
pc <- paircomp(rbind(
  c(1,  1,  1), # a > b, a > c, b > c
  c(1,  1, -1), # a > b, a > c, b < c
  c(1, -1, -1), # a > b, a < c, b < c
  c(1,  1,  1)))

## basic methods
pc
str(pc)
summary(pc)
pc[2:3]
c(pc[2], pc[c(1, 4)])

## methods to extract/set attributes
labels(pc)
labels(pc) <- c("ah", "be", "ce")
pc
mscale(pc)
covariates(pc)
covariates(pc) <- data.frame(foo = factor(c(1, 2, 2), labels = c("foo", "bar")))
covariates(pc)
names(pc)
names(pc) <- LETTERS[1:4]
pc

## reorder() and subset() both select a subset of
## objects and/or reorders the objects
reorder(pc, c("ce", "ah"))


## include paircomp object in a data.frame
## (i.e., with subject covariates)
dat <- data.frame(
  x = rnorm(4),
  y = factor(c(1, 2, 1, 1), labels = c("hansi", "beppi")))
dat$pc <- pc
dat


## formatting with long(er) labels and extended scale
pc2 <- paircomp(rbind(
  c(4,  1,  0),
  c(1,  2, -1),
  c(1, -2, -1),
  c(0,  0,  -3)),
  labels = c("Nordrhein-Westfalen", "Schleswig-Holstein", "Baden-Wuerttemberg"))
## default: abbreviate
print(pc2)
print(pc2, abbreviate = FALSE)
print(pc2, abbreviate = FALSE, width = FALSE)


## paired comparisons with object covariates
pc3 <- paircomp(rbind(
  c(2,  1,  0),
  c(1,  1, -1),
  c(1, -2, -1),
  c(0,  0,  0)),
  labels = c("New York", "Rio", "Tokyo"),
  covariates = data.frame(hemisphere = factor(c(1, 2, 1), labels = c("North", "South"))))
covariates(pc3)